EXPERIMENTAL DESIGNS BALANCED FOR THE ESTIMATION

OF RESIDUAL EFFECTS OF TREATMENTS

[Manuscript received December 2, 19481

Summary
Where an experiment can be carried out by applying different treatments in
succession to the same unit of experimental material, accurate comparisons can be made
between the effects of different treatments. To allow for the residual effect of previous
treatments on the result obtained for any given treatment, it is desirable to adjust the
results for such effects. Methods of constructing balanced designs for the estimation
of these residual effects are described in this paper, and are summarized as follows.
Designs balanced for effect of single preceding treatment : Wnen n , the number of
treatments, is even, a balanced design is possible with n replications ; when n is odd,
2% replications are required.
Designs balanced for the effects of any number of preceding treatments, ignoring the
znteraction of residual effects : When n is a prime or a power of a prime, a balanced
design is possible in n(n-1) replications, which may be set out as a set of n-1 mutually
orthogonal Latin squares, with the same first columns. Designs which are not expressible
a s mutually orthogonal Latin squares are also possible.
Desigm balanced for the eifeat of the two preceding treatments and their interactions :
A design can be developed from a set of n-l mutually orthogonal Latin squares obeying
certain restrictions.
The method of analysis of designs of this type is set out in detail, together with a
numerical example. Direct effects of treatments are shown to be only slightly con-
founded, the maximum confounding being 4 per cent., when there are three treatments.
These designs have u-ide applicability wherever successive treatments can be
applied to the same unit of experimental material.

I. INTRODUCTIOX
Experiments carried out to compare the effects of different treatments on
a given material are discussed in this paper. It deals particularly with those
cases in which the different treatments can be applied in succession to the same
unit of the material. Where this can be done, the treatment effects will be
compared as accurately as possible, since the variation between units will not
affect the comparisons between the effects of treatments. For this accuracy to
be fully realized, however, it is important to make allowance for the effect of
previously applied treatments on the response of the unit of material to a given
treatment.
The estimation of the effects of previous treatments is of direct interest in
experiments, such as crop rotation trials in agriculture, where the aim is to

* Section of Mathematicd Statistics, C.S.I.R.O.

150 E. J. WILLIAMS

determine the succession of treatments (i.e. crops, fallow, etc.) over a number
of years, which will give the best results for the series taken as a whole. This
applies, whether the criterion is yield, market value, or the continued fertility
of the soil. On the other hand, such estimation is of indirect interest only, in
experiments such as those on the feeding of animals. In such cases, while a
group of aninlalx might be tested with a succession of treatments (rations),
the primary interest centres In determining the particular ration which gives
best results according to some criterion. The effect of previous treatments is of
indirect interest, and is estimated only as a correction in determining the effect
of the treatment which follows.
The method of carrying out an experiment by applying successive treatments
to the same unit of material has wide applicability. The main limiting factors
are (i) the length of time taken, compared with simultaneous application of the
different treatments to different units of material, and (ii) the suitability of the
material to repeated treatments. Por example, if the treatment were destructive,
the method would not be applicable.
The unit of experimental material to which successive treatments are
applied will be called a site. The effect of the treatment being applied will be
called the direct effect, and the effect of previous treatment the residual effect,
following Cochran, Autrey, and Cannon(1). We shall discuss the design of
experiments for the estimation of the residual effect, firstly, of the immediately
preceding treatment, and secondly, of the two immediately preceding treatments.
The analysis of such experiments will be discussed and exemplified only for the
former case.
D

(a) Advantages of a Balanced Design

We shall call an experimental arrangement a balanced design when the
direct effect of ccch treatment is associated equally frequently with the residual
effect of each other treatment. Where only the immediately preceding treatment
is to be taken into account in assessing residual effects, it is seen that the condition
for balance is satisfied if each treatment is preceded equally frequently on the
site by each other treatment. Balanced designs have a twofold advantage over
unbalanced designs using the same amount of experimental material ; firstly,
they are more efficient (in the sense that they provide more accurate estimates
both of direct and residual effects), and secondly, the statistical reduction of the
results is relatively simple.
I t is therefore of practical interest and importance to determine how balanced
designs can be constructed, and what; is the minimum number of sites required
for balance. Only those designs in which each treatment is applied once to each
site will be considered. It would, of course, be possible to construct balanced
incomplete block or lattice designs, with the restriction that they were to be
balanced for the estimation of residual effects. Such designs have not been
considered, because, firstly, it seems unlikely that balanced designs of this type
would be required in practice with so many treatments that all could not be
applied at each site ; and secondly, the construction and analysis would be more
complex than for designs with each treatment at each site.
 ESTIMATION O P RESIDUAL EFFECTS 151
The designs given in this paper are balanced for the effect of subsequent
treatment as well as for previous treatment. Such a possibility is not so academic
as at first appears. For example, in carrying out an examination on school
children, the result obtained in one test may well be influenced by the anticipation
of the subsequent tests, to a greater extent, than the actual result of previous
tests.
11. BALANCED DESIGNSFOR A SINGLERESIDUALEFFECT
The conditions for a balanced design are :
(i) Each treatment shall be preceded by each other treatment equaJly
frequently.
(ii) Each treatment shall occur equally frequently at each position, in
order of application to the sites (so that the treatment effects shall be
unaffected by possible effects of order of application).
Each of these conditions implies that the number of sites shall be a multiple
of the number of treatments. If there are n treatments, there are n(n-1)
ordered pairs of treatments. Since n-1 adjacent pairs occur on each site,
condition (i) requires some multiple of n sites' to give all ordered pairs equally
frequently. Condition (ii) is obviously satisfied only by a multiple of n sites.
It will be shown that when n, the number of treatments, is even, balance
can be achieved with a minimum of n replications, and when n is odd, with a
minimum of 2n replications. In fact, balanced designs can be derived from the
cyclic Latin square of side n in which the rows represent the sites, the columns the
order of application, and the symbols the treatments ; two such squares are
required when n is odd. Careful study of Latin squares of sides 3 to 6 has
revealed no minimum solutions other than those derived from the cyclic square,
and it is surmised that no other solutions exist.
Case I : n Even
Let each treatment be represented by one of the classes of residues modulo n
including zero. Condition (i) then can be stated as a requirement that the
differences between the values for two adjacent treatments shall take all values
from 1 to 12-1 equally frequently. If each Werence occurs once in the first;
row of the Latin square, and successive rows 'are obtained by adding 1to &ch
element of the preceding row, the square will fulfil the required conditions. For
each element of any column differs by a constant amount from the corresponding
element of the preceding column, the amount being different for each pair of
adjacent columns. As each treatment occurs once in each column, it will be
preceded on each occasion by a different treatment.
Thus, for example, when n=6, the differences may be taken as
2 5 3 1 4
giving for the Latin square generated the following arrangement :
0 2 1 4 5 3
1 3 2 5 0 4
2 4 3 0 1 5
3 5 4 1 2 0
4 0 5 2 3 1
5 1 0 3 4 2
152 E. J. WILLIAMS

I t is apparent that each treatment is preceded by each other treatment once

and once only.
The problem of determining initial rows of such Latin squares for even
values of .n has the following general solution :
o 1 n-1 2 n-2 3 n-3 ........ 4 2
The row is formed from the numbers 0, n-1, .... in descending order, with
which alternate the numbers 1, 2, .... in ascending order. The successive
differences are :
1 12-2 3 n-4 ............ ;
that is, the odd values in ascending order, alternating with the even values in
descending order, and thus include each difference once only. The Latin square
developed from this initial row is thus balanced for residual effects ofetreatments.
Another method of arriving at a balanced arrangement is applicable whenever
n + l is at prime number. I n this case, the treatments may be represented by
the set of non-zero residue-classes, modulo n + l . The initial row of the Latin
square consists of the residues arranged in numerical order, and successive rows
sre obtained by multiplying the elements of the initial row in turn by 2, 3 ......
to n. Since n + l is prime, each treatment will be represented once only in each
column. Moreover, the differences of successive elements of any row are all the
same as its initial element. Hence, between m y treatment r and the preceding
treatment, every difference, except r , occurs once in the rows which do not begin
with r. Treatment r is therefore preceded once by each other treatment, and
the design is balanced. As an example, n=4 give8
1 2 3 4
2 4 1 3
3 1 4 2
4 3 2 1
At first sight, this method of constructing the balanced square appears to
be different from the method previously given. In the first method each
difference occurs once in each row, wherem in this method, each difference
occurs once between each pair of columns. The two methods can, however, be
shown to be equivalent. For, if x is a residue whose nth power is the lowest
power congruent to 1modulo n + l , then all the powers m, x2, .... LPare different,
and each of the residues 1to n can be &%resented as a power of m. Multiplying
the elements of the first row by the numbers 2 to n to produce the n rows of the
square, is equivalent to adding a constant amount to the index. Hence, if the
treatments are now represented by the indices of m, the differences between the
corresponding elements of adjacent 'columns will be constant. However, as
the design is known to be balanced, it follows that each of the differences from
1to lz-1 must be represented in each row ; where it is to be noted that residues
are now to be taken modulo n and not modulo n f l . More directly, the last
statement follows from the fact that the ratios of successive non-zero residues
are a11 different when the modulus is prime. For example, 2 is a primitive root
of the modulus 5, so that
1=2O
2 =21
3 =23
4 =22
and the above 4-square can be written
0 1 3 3
1 2 0 3
3 0 2 1
2 3 1 0
Another solution, when n is a power of 2, (say 29, results from giving the
successive differences the values
1, 2 . . . . . . . . . .n-1.
Any cumulative sum of successive differences is of the form
a---
( a + l ) b(b3.1)
2 a
reduced modulo 25 where a and b both take values between 0 and 2"l. Since
the expression equals (a-b)(af b+li , and neither a-b nor a + b + l is a
2
multiple of 2h1, the expression is not a multiple of 2% Hence, no cumulative
sum of successive differences is zero, and these differences provide a solution.

INITIAL ROWS FOR BALANCED SQUARES : 12 EVEN

Other sets of suitable initial rows have been determined by trial, and are
given for n=4, 6, 8, in Table 1. When one initial row has been found, others
can be derived from it by multiplying each element by a residue which is prime
to rz, and adding an arbitrary integer. Such rows are equivalent, since the
effect of multiplication is the same as permutation of the numbers representing
the treatments, and a rearrangement of the rows, and the addition simply
permutes the rows. Only a standard representative of each such set of rows is
therefore tabulated.
Case I1 : n Odd
When n is odd, the sum of the residues is a multiple of n. Hence, if every
difference were to occur once in a row, the last element would be the same as
the first. It can be seen, therefore, that balanced designs based on a single
cyclic Latin square are impossible. I t is, however, possible to achieve balance
with a pair of such squares, the differences not represented in one square being
represented twice in the other.
For example, when n=5, a possible pair of sets of differences, generating
s pair of Latin squapes which together provide balance, is
1 2 1 3 ' 2 4 3 4
154 E. J . WILLIAMS

and the corresponding Latin squares are

0 1 3 4 2 0 2 1 4 3
1 2 4 0 3 1 3 2 0 4
2 3 0 1 4 2 4 3 1 0
3 4 1 2 0 3 0 4 2 1
4 0 2 3 1 4 1 0 3 2
It is easy to prove that a pair of Latin squares can be found to give balance
for every odd value of n. The initial row
n-1 -
- n+l
0 1 n-1 2 n-2 ...... 2 2
gives each odd difference twice, and the reverse arrangement
n+l n-1
- - . . . . . . 12-2 2 12-1 1 0
2 2
gives each even difference twice. The pair of Latin squares generated from
this pair of initial rows therefore provides a balanced design.
Another form of general solution is possible when n is a prime of the form
4m+3. The set of differences
1 2 . . . . . . . . . . . . . . 2m+l 2m+l . . . . . . . . . . . . . . 2 1
and the opposite set,
4mf2 4mfl . . . . . . 2m+2 2 m +2 . . . . . . 4 m +l 4 m +2
together include every difference twice. The rows generated by one set give,
on reversal, the rows generated by the other. If the second set of differences
is doubled, it becomes
4 w + l 4m-1 . . . . . . 3 1 1 3 . . . . . . 4m-1 4m+l
If the middle element of the initial row is taken as zero, the elements to the
right are the quadratic residues (modulo 4m+3), and the elements to the left
are their opposites. Hence every residue occurs but once in the row, and the
two sets of differences generate a, balanced design.

INITIAL ROWS FOR BALANCED PAIRS O F SQUARES : 12 ODD

Paire of initial rows are given, for m=3, 5 , 7 in Table 2. As for even values
of lz, other pairs may be derived from these by multiplying each element by a
residue prime to n, and adding an arbitrary integer.
 ESTLMATION O F RESIDUAL EFFECTS 155

It is of both practical and theoretical interest to investigate designs which

will give balance for the effects of the two immediately preceding treatments.
I n deciding on the requirements for these designs, two possibilities arise ; we
may be prepared to ignore the interaction of one preceding treatment on the
other, in assessing their effect on the third, or we may wish to allow for their
interaction as well.
I n the former case, each treatment must be immediately preceded by each
other treatment equally frequently, and must also be preceded by each other
treatment by two positions equally frequently. The former condition, as was
shown above, requires a multiple of n replications. As in any set of n replica-
tions, any treatment occupies the first and second positions in order on the site
once each, other treatments can occupy the position two before that treatment
in only n -2 of the replications. Therefore, in order that each of the n -1 other
treatments precede a given treatment by two positions, a multiple of ra-1
replications is required. From this it follows that only multiples of n(n-1)
replications are admissible, n ( n - 1 ) being the minimum possible.
I n the ~econdcase, when interactions are not ignored, it can be seen that
each ordered trio of treatments must occur together equally frequently. Noy
there are n ( n - l ) ( n - 2 ) such ordered trios, and n-2 occur in each replication.
Hence a multiple of n ( n-1) replications is required in this case also.
The latter case, where possible interactions are not ignored, is practically
the more useful. I t is, however, more restrictive, for designs satisfying the
former set of requirements can be found which do not satisfy the latter. The
following trio of 4 x 4 Latin squares illustrates this.
0 1 3 2 0 3 2 1 0 2 1 3
1 2 0 3 1 2 3 0 1 0 3 2
2 3 1 0 2 0 1 3 2 3 0 1
3. 0 2 1 3 1 0 2 3 1 2 0
It is easily seen that each treatment is preceded by each other treatment, and
also preceded by two positions by each other treatment ; but, of the
4 x 3 x 2 =twenty-four possible ordered trios, six occur twice, twelve occur once,
and six not at all.
If designs satisfying the more stringent conditions can be constructed with
just n ( n - 1 ) replications, it will be advantageous to use such designs rather
than the less restricted ones.
It has been proved(2, 3) that, whenever n is a prime number or a power of s
prime, a completely orthogonalizad square of side n can be constructed. A
method of forming a design in n(n-1) replioations, balanced for pairs of residual
effects and their interaction will be illustrated, using the properties of the com-
pletely orthogonalized square. I t is believed that such a design can be
constructed whenever a completely orthogonalized square is possible, but we
have not been able to prove this. I t seems unlikely that such a design in %(a-1)
replications can be constructed other than by such means, such designs being,
156 E. J. WILLIAMS

therefore, available only when n is a prime or a power of a prime.* This is no

great disadvantage, however, as, of the numbers from 3 to 9, the likely practicable
range, only one is not a power of a prime. I t would be no great inconvenience
in such cases to add another treatment, or to duplicate one of the existing ones,
to bring the number up to a prime power.
I t will first be shown that any set of ?z -1 mutually orthogonal Latin squares
gives a design balanced for the effect of any number of preceding treatments,
when their interactions are not taken into account, provided that the initial
column of each square (or, in fact, any column, the same for each square) has
the elements in the same order. This having been proved, it is necessary only
to discuss the conditions under which such a set of orthogonal squares is balanced
for interactions as well.
Any set of n-1 mutually orthogonal Latin squares may be represented as
&.set of .n classifications (vie. those corresponding to the rows, and to the symbols
of the .n-1 squares), which are mutually orthogonal, and orthogonal with
columns.
The initial columns of each square may be made identical by suitably
numbering the elements of the successive squares, and the rows will be numbered
in the same way as the initial columns.
The corresponding symbols for a given column in different squares may
now be written down in order, preceded in each case by their row symbol. I n
this way, the n columns will give a set of n squares of side n. The elements of
the first square will be in the same order in each column (since they are the
initial columns of the original squares). It will be proved that the remaining
n -1 squares are mutually orthogonal Latin squares.
To the original set of Latin squares corresponds a balanced incomplete
block design (r = n + l , v =n2, b =?t2+n, k =n, h =l). There are n f l complete
replications, each comprising n blocks of n elements. The first replication may
be taken to correspond with the classification of the n2 elements by columns.
The n elements in the first column will be allocated to different blocks in each
of the remaining n replications. We shall number the blocks in which a given
element of the first column occurs, the same in each of these n replications. This
is equivalent to making the initial columns of the original orthogonal squares
identical. In each such replication, the n elements of any column, other than
the first, must occur in a differently numbered block. Also any given element
of the column must occur in a differently numbered block in each such replication,
in order that it may be associated once in the same block with each element
of the first column. Hence the square of block numbers, classified according
to element of column, and replication, is a Latin square. Furthermore, the
elements of any column must be associated once and once only with the elements
of each other column. Hence the ?t -1 Latin squares corresponding to the .n-1

* Since this paper went to press, designs based on sets of oyclic Latin squares which are not
orthogonal have been found, including designs for n=6 and n= 10. I t is intended t o describe
these in a subsequent note.
 ESTIMATIOX OF RESIDUAL EFFECTS 157

columns, are mutually orthogonal. The n classifkations, corresponding to the

symbols of the n squares, are therefore mutually orthogonal.
The formation of this set of orthogonal squares by interchanging columns
and squares is illustrated below, for the case n=4.

Classification
Row Symbols Square A Square B Square C
Column .. I I1 111 I V I I1 111 I V I I1 I11 I V I I1 I11 I V
1 1 1 1 1 4 2 3 1 3 4 2 1 2 3 4
2 2 2 2 2 3 1 4 2 4 3 1 2 1 4 3
3 3 3 3 3 2 4 1 3 1 2 4 3 4 1 2
4 4 4 4 4 1 3 2 4 2 1 3 4 3 2 1

Corresponding columns of different squares :

Column I Column I1 Column I11 Column IV
Classification R A B C R A B C R A B C R A B C
1 1 1 1 1 4 3 2 1 2 4 3 1 3 2 4
2 2 2 2 2 3 4 1 2 1 3 4 2 4 1 3
3 3 3 3 3 2 1 4 3 4 2 1 3 1 4 2
4 4 4 4 4 1 2 3 4 3 1 2 4 2 3 1

Since any pair of columns gives rise to an orthogonal pair of classifications

in this way, corresponding elements of any pair of columns give every ordered
pair of elements (including twins) once. Twins arise only from the row symbols,
which are the same for each column. Hence, when these are disregasded, every
element is associated with every other element onoe, in any pair of columns.
This establishes the fact that the original set of orthogonal Latin squares gives
a balanced design for the effect of any preceding treatments whatever, and
consequently for any number of preceding treatments, when their effects are
regarded as additive.
There are other possible solutions, however. The illustration given above
is an example of this, for it is not a set of orthogonal squares, and cannot be
transformed into one by permutation of the rows.

(a) Bets of n-1 Mutually Orthogonal Lati% Bquares giving Balanced

Designs for Two Preceding Treatments and Their Interaction
It was shown above that, when interactions are taken into account, a design
is required in which every ordered trio occurs equally frequently. The cases
when n is a prime or a power of a prime can both be treated in the same way ;
but as the former case is the simpler, and for it a general solution has been found,
it will be treated first. For the case when n is a power of a prime, use is made
of the properties of Galois fields.

Case I : n an Odd Prime Number

As before, each treatment is represented by one of the classes of residues,
modulo n , including zero. These classes are the elements of a field, since n is
prime. Suppose that a set of m-1 non-zero merences (not necesssbsily all
distinct) between these elements can be formed, such that (i) no cumulative
158 E. J. WILLIAMS

sum of successive differences is zero, and (ii)the ratios of consecutive differences

take all values except 0, 12-1.
It is noted that, as a result of condition (ii),the first asd last differences are
equal ; for their ratio is the product of all non-zero residues other than n -1,
which is congruent to 1.
If the set of differences be multiplied by 1,2, . . . . . . lz -1, in turn, n -1 sets
will be generated, which will comprise between them equally frequently, every
non-zero element of the field. That this is so follows from the fact that, .n
.
being prime, any element multiplied in turn by 1, 2, . . . . . n -1, generates
every non-zero element. Moreover, since every such element occurs at each
position in the set, in one or other of the sets, it will bear to each difference that
precedes it a different ratio (other than n-I), SO that each will be preceded by
each other difference except its opposite, exactly once. In other words, each
ordered pair of differences, except those whose sum is zero, will occur once in
some set.
Corresponding to each set of differences, a Latin square can be formed, as
shown above. The lz-1 Latin squares so produced are dearly mutually
orthogonal cyclic squares, as their mode of formation is equivalent to that
described by Stevens(3). Since every admissible ordered pair of differences is
adjoined to each element by this means, in one or other of the squares, every
ordered trio of elements is thereby produced. The set of Latin squares therefore
gives the required balanced design.
The method of formation is made clear by means of an example, for n=5.
The set of differences
1 3 3 1
satisfies the conditions (i) and (ii) above ; the successive ratios are
3 1 2
and the ratio 4 is inadmissible.
Three other sets of differences may be produced by multiplying the given
set by 2, 3, and 4, successively ; the four sets are
1 3 3 1
2 1 1 2
3 4 4 3
4 2 2 4
From these four sets, the following four orthogonal Latin squares are derived.
Every ordered trio is seen to be represented once.
0 1 4 2 3 0 2 3 4 1
1 2 0 3 4 1 3 4 0 2
2 3 1 4 0 2 4 0 1 3
3 4 2 0 1 3 0 1 2 4
4 0 3 1 2 4 1 2 3 0
 ESTIMATION OF RESIDUAL EFFECTS 159
For any odd prime number, a set of differences satisfying conditions
(i) and (ii) is

and the corresponding initial row of the first Latin square of the set is

The m(n-1) rows obtained by multiplying the elements of the initial row by
1, 2, 3, . . . . . . n-1, and in each case adding to the resulting elements
0, 1,2, . . . . . . n-1 give the required design balanced for pairs of residual effects.
This proves the existence of such a design when n is any odd prime.
This general solution is the same as that found above for a single Latin
square giving balance for the effect of one preceding treatment only, when
n is even, or for the pair of Latin squares required when n is odd.

DESIGNS BALANCED FOR PAIRS OF RESIDUAL EFFECTS.

INITIAL ROWS FOR GENERATING SETS OF ORTHOGONAL
SQUARES. n A PRIME

Other initial rows also give balanced designs in the same way ; these are
tabulated in Table 3, for 12=3, 5, 7.
Case I1 : n a Power of a Prime
When n=pk, where p is prime and 3c an integer, each treatment may be
represented by one of the elements of a Galois field. Galois fields exist only for rz
a prime or a power of a prime. A full treatment of the properties of these fields
will be found in Carmichael (4, pp. 242-88).
Every field has a primitive element a, with the property that the equation
m m = l is satisfied only by m a multiple of n-1. Consequently, the elements
x, a2, x3, . . . . . . an-l (=I)are all distinct, and therefore include every non-zero
element of the field.
It can be shown that such s primitive element satisfies an equation
P(m)=0, where P(x) is an irreducible polynomial of degree 5, whose coefficients
are members of GP(p). P ( m ) is called a minimum function. It follows that
every element of the field may be expressed as a polynomial in x of degree k-1,
namely
+
c,, +c1x . . . . . . +C k-lmkl,
160 E. J. WILLIAMS

. ...
where the coefficients c,, c,, . . c,-, are members of Gl?(p). Addition and
multiplication are defined as ordinary addition and multiplication, followed by
simultaneous reduction modulo p and modulo l?(x).
We note that, when n is the power of an odd prime,
xn-1-1 =O

but

therefore -

When n is a power of 2,
1+1=o.
We are now in a position to adapt the procedure used when n is prime, to
the derivation of a balanced design when .n is a power of a prime. We require
to find a set of 12-1 non-zero differences among the elements of the field, such
that (i) no cumulative sum of successive differences is 0, and (ii) the ratios of
12-1
successive differences take all values except 0, x 7 (when n is odd) and all values
except 0, 1 (when is a power of 2).
As for the case of n a prime, the first and lest differences must be equal.
..
If the set of differences be multiplied by 1, x, x2, . . . . in turn, n -1
sets will be generated. Corresponding to each such set of differences, a Latin
square can be formed. The argument given above for the case when n is prime
can be extended to show that the Latin squares so constructed are mutually
orthogonal, and that every ordered trio of elements is represented once. This
set of squares therefore gives the required balanced design.
Although it is believed that a Bolution exists for every n which is a power of
a prime, no general solution has been found. Solutions for the particular cases
likely to be required in practice, namely, n=4, 8, 9, are, however, presented in
Table 4. The initial row is given, also the minimum function and the power
cycle for the primitive root x. The initial row is standardized by choosing 0 and 1 -
as the first two elements.
From any given solution, other solutions may be derived by raising each
element to a power whose index is a power of p. The addition and multiplication
relationships among the new elements are exactly the same as those among the
old. For the pth power of a sum is congruent (modulo p ) to the sum of the
pth powers ; while the pth power of a product is obviously identical with the
product of pth powers. There are just k different solutions so derivable ; for
if x is any element,
$9 =x" =x.
This fact enables initial rows to be further standardized. For example, when
n=9, the following two initial rows are equivalent :
0 1 x $5 x2 $3 $4 $7 x6
0 1 x3 m7 x6 x rC4 x5 x2
In Table 4, the solution is given for which the third element is the lowest possible
power of x.
 ESTIMATION O F RESIDUAL EFFECTS 161
As an example of the generation of a balanced design from the initial row

.
given in Table 4, the complete design. is given for n =4. I n setting out the design
in this way, it-is convenient to represent the zero element by 0, the unit element
..
by 1, and the successive powers of m by 2, 3, . . . . n-1.
0 1 2 3 0 2 3 1 0 3 1 2
1 0 3 2 1 3 2 0 1 2 0 3
2 3 0 1 2 0 1 3 2 1 3 0
3 2 1 0 3 1 0 2 3 0 2 1

DESIGNS BALANCED FOR THE EFFECTS OF PALRS OF RESIDUAL EFBECTS. INITIAL ROWS FOR
. GENERATING SETS OF ORTHOGONAL SQUARES WHEN ?% IS A POWER OF A PRIME

12

-.-
Initial ROW 1 fE2E
I
I

I
Power Cycles

x=x, x2=x+l, xS=l

IV. ~NALYSIS OF RESULTS,TAKINGINTO AOCOUKT RESIDUALEFFECT

OF PRECEDING TREATMENT
There are four factors to be taken into account in the analysis ; sites or
replications, position in order of application to the site, and direct and residual
effects of treatments. It is noted first of all that, since each treatment occurs
equally frequently at each position, positions are orthogonal with each of the
other effects. The determination of the effect of the positions therefore is made
direct from their means without adjustment. Sites and direct effects are also
orthogonal, but neither is orthogonal with residual effects. The analysis ignoring
residual effects is straightforward, and follows the lines of the analysis of a
Latin square or set of Latin squares. When residual effects are taken into
account, the analysis is more complex, but as the design is balanced, the com-
putations are not difficult.
I t is useful to note that sites bear the same relationship to residual effects
as do direct effects. For consider all the sites in which treatment i occupies the
final position. On these sites, the residual effect of treatment i alone is absent.
Comparisons among such sites, in which the same treatment comes last, are
therefore orthogonal with residual effects of treatments. On the other hand,
comparisons between the groups of sites with different final treatments are not
 E. J. WILLIAMS

orthogonal with residual effects. Since treatment i is never preceded by treat-

ment i, the total for any treatment is affected by residual effects in exactly the
same way as is the total for any set of replications for which that treatment
comes last. U

(a) hTotationfor the Description of the Analysis

'n =number of treatments,
mn =number of sites or replications,
Si,=total of results for the kth site of those for which treat-
ment i is the final one,
Pi =total of results for all sites for which treatment i is t h e
Totals final one (designated group i),
T i =total of results for treatment i,
R , =total of results for treatments preceded by treatment i,
Pj =total of results for position j in order of application to
the site,
-G =grand total of all results.

Estimates (f, =average effect for all sites of group i,

(measured as ti =direct effect of treatment i,
r i =residual effect of treatment i,
from zero) r r i =residual effect of treatment i (direct effects beingignored),
l p j =effect of position, j in order of application.

( b ) Derivation of the Estimates, and Tests of Significance

The normal equations are
G
m(nf, -ri) =Pi- -
12. '

The solutions of the normal equations are

 ESTIMATION O F RESIDUAL EFFECTS

p =A.
nP.-G
mn2
The estimates given here take into account only the information given by
comparisons within sites (actually, within groups of sites with the same final
treatment). If the information provided by comparisons between groups of
sites be taken into account, slightly more accurate estimates result.
It can be shown that residual effects are confounded 1/(n2--n)with groups
of sites and with treatments. Since only that part confounded with sites is
potentially recoverable, the fraction of recoverable information confounded with
sites is l / ( m 2 -n -1).
I n the same way it is found that direct effects of treatments are confounded
l / ( n 2-n) with residual effects when sites are not taken into account. The fraction
thus confounded is of course not recoverable, so that the fraction potentially
recoverable is (n2-.n -1)/(n2 -n). When sites are taken into account, an
additional fraction l / ( ( n 2-n)(n2 -n -1)) is confounded. Hence, the fraction of
potentially recoverable information on direct effects, confounded with sites, is
l / ( n 2-n The efficiency factors for estimates from comparisons within
sites are therefore as follows :

Residual effects

EFFICIENCY FACTOR O F ESTIMATES OF DIRECT AxD RESIDUAL

EFFECTS

1 Residual

These factors are presented, for small values of ?z, in Table 5. Unless n is small,
the gain due to recovery of information from comparisons among sites will be
slight, and will moreover be offset to some extent by the loss of information due
to inaccuracies in weight,ing. However, recovery of information on residual
effects may sometimes be worth while, when n is 3 or 4.
I n many cases, where the estimation of the residual effect is not of direct
interest, and it is required only as an adjustment to the average effect of a
164 E. J. WILLIAMS

treatment, estimates of the residual effect with improved accuracy are irrelevant.
The recovery of information Is, however, discussed below, to meet the oases in
which the residual effects are of direct interest.
Ordinarily, two analyses of variance will be required for testing the
significance of treatments and of the djustment for residual effects. The sum
of squares for direct effects, corrected for residual effects, and the sum of squares
for residual effects, corrected for direct effects, will both need to be calculated.
The two analyses are set out in Table 6.

Sum of Squares

Degrees of
Freedom Direct Effects Residual Effects
Adjusted / Adjusted

Sites-Between groups

Within groups ..

Positions .. ..

Direct effects .. ..

Residual effects ..

Error .. .. .. by subtraction

Total

( c ) Btandard Errors of Estimates of Treatment Effects

The estimated variance of the differelm between two estimates of direct
effects, when residual effects are ignored, is

where E is the error mean square from the analysis of variance. When the
adjustment for residual effects is made, the variance of a difference is increased to
 ESTIMATION OF RESIDUAL EFFEOTS 165
The variance of the difference between the estimates for two residual
effects is
2nE
m(n2-n -2) '
( d ) Recovery of Information between Groups of Bites
Let the error variance be 1/W, and the variance between sites of the same
group, on a single-result basis, be l/W'.
I f E is the error mean square and S the mean square between sites within
groups, their expected values are respectively 1 / W and l/Wr. However, if E
exceeds B, we assume W and W' to be equal, and take 1/E as the eatimate of
each.
The estimate of the residual effect of treatment, after including information
from groups of sites, is

The estimated variance of the difference between two such estimates is

The expression for the variance of differences may be compared with that obtained
above, without recovery of information. The expression in brackets represents
the reciprocal of the relative efficiency with recovery of information.

ORDER OF APPLICATION OF CONCEYTRATIOYS IN

SUCCESSIVE REPLICATIONS

Replication

Numerical Example
The example given below is taken from experimental pulp evaluation
technique. Samples of pulp suspension at varying concentrations are beaten
in a Lamp& mill, to determine the effect of concentration on the properties of
the resulting sheets. Observations of the condition of the mill after each beating
indicate that certain concentrations of pulp have m effect on the mill which may
166 E. J. WILLIAMS

affect the next besting. I n order to allow for this effect if it exists, a balanced
design is adopted. There are six concentrations used, in all, and six replications
are carried out. The layout of the experiment is shown in Table 7 and the results
obtained for the burst factor of the resulting sheets in Table 8.

TABLE8
BURST FACTOR RESULTS F O R THE EXPERIXENT O F TABLE 7

Replication Total

Total

TABLE9
TOTALS AND SUMS O F S Q U m E S REQLTRED FOR ANALYSIS

In terms of
estimates

In terms of
totals
i

1
2
3
4
5
6

Add P ,

I
I
Divisor for sum of squares .. , 216
Sum of squares . . .. .. 69.298 24.076 1 23.436
I
 ESTIi?IATION O F RESIDUAL EFFECTS 167
For numerical work it is convenient to use the quantities given below in cal-
culating the estimates and the sums of squares in the analysis of variance.
Corresponding to any treatment, all but the first of these quantities may be

TABLE10
ANALYSIS O F VARIANCE

Sum of Squares
Degrees
of Mean Square
Freedom Direct Effects Residual Effects (Adjusted)
Adjusted Adjusted

Replications ..
Positions in order
of beating ..
Direct effects . .
Residual effects
Error .. ..

Total .. 165.062 I
1
I

** Significant a t the 1% level.

calculated successively on the machine without clearing out the previously

obtained total. In the final column is shown the divisor for the sum of squares
of the quantities in the previous column, to give the appropriate sum of squares
for the analysis of variance.

TABLE11
ADJUSTED TREATMENT MEANS AND ESTIMATES O F RESIDUAL EFFECTS

I .
Adjusted Residual
Treatment 1 Mean Effect

1
2
3
4
5
6
Standard error
of difference 0.46
168 E. J. WILLIAMS

Divisor
A =mn2t', =nT -G ma3
B i =nan(n2-n-l)r', =n2Ri +%Pi +nPl -(n +1)G rnn3(n2-n -1)
C i =mn(n2 -n -2)ri =n2Ri +nPi+%Ti +nPl - ( n f 2 ) G mn3(n2-n - 2 )
=B, + A i
D , =mw(n2-n -2)ti =(n2 -n -1)l', +nR, + P i +PI -nG mn(n2-n -1)(n2 -n -2)

The detailed analysis of the results of the experiment is set out in Tables
9, 10, and 11.

V. ACKNOWLEDGMENT
The work described in this paper was carried out as part of the research
programme of the Section of Mathematical Statistics, 0.8.I.R.O.

VI. REFERENCES
(1) COCHRAN, W. G., AUTREY, K. M., and CANNON, C. Y.-A double change-over design for dairy
cattle feeding experiments. J. Dairy Sci. 24 : 937 (1941).
(2) B o s ~ R.
, C.-On the application of the properties of Galois fields to the problem of con-
struction of Hyper-Graeco-Latin squares. Sankhyd 3 : 323 (1938).
(3) STEVENS, W. L.-The completely orthogonalized Latin square. Ann. Eugen. 9 : 82 (1939).
(4) CAR~CHAEL, R. D.-" Introduction to the Theory of Groups of Finite Order." (Ginn and
Co. : Boston and London, 1937.)